16th World Congress of the International Fuzzy Systems Association (IFSA)

9th Conference of the European Society for Fuzzy Logic and Technology (EUSFLAT)

Weighted Average Operators Generated by

n-dimensional Overlaps and an Application in Decision
Making
Ivanosca A. da Silva1 Benjamín Bedregal2 Humberto Bustince3
1
Instituto Metrópole Digital, Universidade Federal do Rio Grande do Norte
2
Departamento de Informática e Matemática Aplicada, Universidade Federal do Rio Grande do Norte
3
Departamento de Automática y Computación, Universidad Pública de Navarra

Abstract This paper proposes a generalization of the weighted

average operator, by considering a n-dimensional over-
In this paper we provide a way to generate a class of lap function instead of the summatory. It is proved
weighted average operator from n-dimensional overlap that the resulting operator results also in an aggrega-
functions and aggregation functions. These weighted tion operator and we apply one of this overlap generat-
average operators are used in an algorithm of a multi- ing weighted operator in an algorithm of multi-attribute
attribute group decision making problem based on deci- group decision making based on decision matrices. The
sion matrix. An illustrative example is considered with paper is organized as follows: In section 3 we consider
the application of our method for two specific weighted the notion of n-dimensional overlap functions and a way
average operators of this class. This result is compared to obtain n-dimensional overlap functions from overlap
with the ranking of other methods. functions. Section 3 presents how to obtain an average
operator from arbitrary overlaps and an algorithm to ap-
Keywords: n-dimensional overlaps, weighted average, ply this operator in Multi-attribute Group Decision Mak-
multi-attribute group decision making ing based on decision matrix. Finally section 4 presents
the final considerations and future works.
1. Introduction
2. n-Dimensional Overlap Functions
Bustince et al. in [4] introduced a new class of
Overlap functions are special kinds of non-necessarily
aggregation functions called overlap functions, which
associative aggregation operators proposed in [4] in or-
are basically continuous, positive and commutative 2-
der to be applied in classification problems involving the
dimensional aggregation operators [3]. From them sev-
overlap problem and when the associativity property is
eral theoretical and applied research on this kind of func-
not strongly required, as in image processing and deci-
tion have been made (see for example [2, 6, 10, 11, 12,
sion making based on aggregation operators [13].
17, 19]). Overlap function has been applied in classifica-
tion problems where the classes are not clearly separated
Definition 2.1 A bivariate function O : [0, 1]2 → [0, 1]
(see for example [16]) and in decision making based on
is said to be an overlap function if it satisfies the follow-
fuzzy preference relations (see for example [6]).
ing conditions:
In [16], Gómez et al. extend the notion of overlap
function for n-dimensional overlaps in order to measure (O1) O is symmetric;
the degree of overlapping of several classes in classifica-
tion systems. They also study properties of migrativity, (O2) O(x, y) = 0 if and only if xy = 0;
homogeneity and Lipschitz continuity for n-dimensional
overlaps. In addition it was also given an example in (O3) O(x, y) = 1 if and only if xy = 1;
classification problems using Fuzzy Rule-Based Classi- (O4) O is non-decreasing;
fication Systems where the use of n-dimensional overlap
functions provide a better result than the product t-norm, (O5) O is continuous.
which is the usual in this kind of problems.
On the other hand, Multi-attribute Group Decision Examples of overlap functions are the continuous
Making consists in a choice of one or more alternatives t-norms with no zero divisors (property (O2)) and
among several ones by a group of decision makers (ex- Op (x, y) = xp y p , with 0 < p, which is not a t-norm
perts) who provide a matrix to express how much each when p 6= 1 [2].
alternative satisfies each one of the attributes considered It is well known that associative overlaps are positive
[5, 15, 14, 22, 23]. Fuzzy logic, by its nature, has played continuous t-norms [4, 10]. On the other hand, the as-
an important role in the field of decision making, since it sociativity property of the t-norms allows to extend each
is usual that decision makers are subject to some uncer- t-norm in a unique way to a n-dimensional operation as
tainty, which can be expressed in terms of fuzzy degrees. follows [18, Remark 1.10.(i)]:

© 2015. The authors - Published by Atlantis Press 1473

 Proof: Straightforward. 2
Notice that when O is associative, i.e. is a t-norm, and
T (x1 , . . . , xn ) = T (x1 , T (x2 , . . . , T (xn−1 , xn ) · · · ) A is idempotent, i.e. is an average aggregation operator,
n
then OA = On . In fact, in general, it is desirable that A
The unicity referred by Klement, is an average aggregation function.
Mesiar and Pap in [18], is in the sense
that T (x1 , T (x2 , . . . , T (xn−1 , xn ) · · · ) = Example 2.1 By considering O2 and the maximum ag-
T (T (. . . T (T (x1 , x2 ), . . . , xn−1 ), xn ). However, gregation operator M we obtain the average operator
for the case of overlaps which can not be associative,
n
this is not possible. In particular, considering Op for n
Y
p = 2 and n = 3, O2 (0.4, O2 (0.5, 0.8)) = 0.163 6= O2M (x1 , . . . , xn ) = O2n (x(1) , . . . , x(n) ) = x2i
(i)
0.162 ∗ 0.25 = O2 (O2 (0.4, 0.5), 0.8). i=1

On the other hand, Gómez et al. in [16], extended the where (x(1) , . . . , x(n) ) is the permutation of
notion of overlaps for n-dimensional overlaps as seen in (x1 , . . . , xn ) such that x(i) ≥ x(i+1) for each
the following definition: i = 1, . . . , n − 1.
On the other hand,
Definition 2.2 A function O : [0, 1]n → [0, 1] is said
to be a n-dimensional overlap function if it satisfies the n
Y 1
n n
following conditions: O0.5M (x1 , . . . , xn ) = O0.5 (x(1) , . . . , x(n) ) = x(i)
2i

i=1
(O1) O is symmetric;
n
where (x(1) , . . . , x(n) ) is the permutation of
(x1 , . . . , xn ) such that x(i) ≤ x(i+1) for each
Q
(O2) O(x1 , . . . , xn ) = 0 if and only if xi = 0;
i=1 i = 1, . . . , n − 1.
n
Q
(O3) O(x1 , . . . , xn ) = 1 if and only if xi = 1; Notice that there are n-dimensional overlaps which
i=1
are not generated this way. For example, the Einstein
(O4) O is non-decreasing; product aggregation operator defined in [16] by

(O5) O is continuous. n
Q
xi
i=1
A n-dimensional overlap is said strict if also satisfies EP (x1 , . . . , xn ) = n
Q
the property 1+ (1 − xi )
i=1
(O6) O(x1 , x2 , . . . , xn ) < O(y, x2 , . . . , xn ) when
Qn is a n-dimensional overlap which cannot be generated
xi > 0; and x1 < y. from an overlap and an aggregation operator according
i=1
to Eq. (2).
Let O be an overlap and n a positive natural number.
Define the function, On : [0, 1]n → [0, 1] given by 3. Application of n-dimensional overlaps in multiple
attribute group decision making
On (x1 , . . . , xn ) = O(x1 , O(x2 , . . . , O(xn−1 , xn ) · · · )
(1) The weighted average (WA) and some variants of it are
Observe that On is not an overlap when n ≥ 3 because the most applied aggregation operators found in the liter-
the symmetry fails. The next theorem intends to recover ature. For example, they have been used in a wide range
the symmetry for On . of different subjects such as statistics, economics and
engineering [20]. In particular, there are several meth-
Theorem 2.1 Let O be an overlap, n a positive natu- ods in decision making which consider a fuzzy or an
ral number and A : [0, 1]n! → [0, 1] be a continuous extension of fuzzy framework that uses aggregation op-
aggregation function such that erators like the weighted average and ordered weighted
(A1) if A(x1 , . . . , xn! ) = 1 then xi = 1 for some i = average. For example, [7, 8] present a method which
1, . . . , n! uses the generalization of the weighted average operator
proposed by their respectives authors based on interval-
(A2) if A(x1 , . . . , xn! ) = 0 then xi = 0 for some i = valued intuitionistic fuzzy values [1].
1, . . . , n!.
n 3.1. Weighted average operators based on
Then the function OA : [0, 1]n → [0, 1] defined by
n-dimensional overlaps
n
OA (~x) = A(On (~x(1) ), . . . , On (~x(n!) )) (2) Let F be a n-dimensional function satisfying (O2) and
is a n-dimensional overlap function when ~x(i) is the i-th w = (w1 , . . . , wn )T be a positive weighted vector, i.e.
n
permutation1 of ~x ∈ [0, 1]n
P
wi 6= 0 for each i = 1, . . . , n and wi = 1. The
i=1
1 for any n-tuple of values there are just n! possible permutations. weighted average operator based on F and the positive

1474
weighted vector w is the mapping W AF n
w : [0, 1] → the attributes. For that, each decision maker dk deter-
(k)
[0, 1] defined by mines a decision matrix M (k) = (mij )n×m , where the
rows represent the alternatives and the columns repre-
sent the attributes. In classic decision making, accord-
F (w1 x1 , . . . , wn xn )
W AF
w (x1 , . . . , xn ) = (3) ing to the opinion of the decision makers dk , the position
F (w1 , . . . , wn ) (k)
mij of M (k) has the value 1, in case the alternative xi
Notice that if F is the Łukasiewicz extended t- has the attribute aj and zero otherwise. Nevertheless, in
n
P
conorm, i.e. F (x1 , . . . , xn ) = min{ xi , 1}, then several situations, some attributes are inherently fuzzy,
i=1 for example “quality of construction project”, and so it
W AF w is the usual weighted average operator for pos- must be dealt as a fuzzy set. In this case, the value in
(k)
itive weighted vectors. the position mij would be the membership degree, i.e.
In the following we will consider the case when F is a value in [0, 1], of the alternative xi to the fuzzy set
a n-dimensional overlap. associated to the attribute aj . In general, we have two
types of attributes: benefit and cost. For example, the
Theorem 3.1 Let O be a n-dimensional overlap and
attribute “quality of construction project” is a benefit at-
w = (w1 , . . . , wn )T be a positive weighted vector. Then
tribute whereas “risk of investment” is a cost attribute.
W AO w is a continuous n-ary aggregation function satis- Let I be the set of index of the benefit attributes.
fying the property (O2). In addition, if O satisfies (O6)
We propose the following solution for MAGDMP:
then W AO w satisfies (O3).

Proof: Since the product and On are increasing, Step 1 Transform each decision matrix M (k) into the
(k)
it is clear that W AO w is also increasing. Moreover, standard decision matrix R(k) = (rij )n×m as fol-
O(0,...,0)
W AO w (0, . . . , 0) = 0
O(w1 ,...,wn ) = O(w1 ,...,wn ) = 0
lows:
O(w1 ,...,wn )
and W AO w (1, . . . , 1) = O(w1 ,...,wn ) = 1. Notice that (
(k)
O(w1 , . . . , wn ) 6= 0 because each wi 6= 0. There- (k) mij if j ∈ I
rij = (4)
fore, W AO w is a continuous n-ary aggregation oper-
(k)
1 − mij if j 6∈ I
ator satisfying (O2). Moreover, suppose that O sat-
isfies (O6). If W AO w (x1 , . . . , xn ) = 1 then, by Step 2 Given an n-dimensional overlap O, in order to
(O2), we have that xi 6= 0 for each i = 1, . . . , n obtain a collective decision matrix C = (cij )n×m ,
and by Eq. (3), O(w 1 x1 ,...,wn xn )
O(w1 ,...,wn ) = 1 and therefore aggregate the standard decision matrices with
O(w1 x1 , . . . , wn xn ) = O(w1 , . . . , wn ). Since wi xi ≤ W AOω as follows:
wi for each i = 1, . . . , n, if for some i = 1, . . . , n,
xi 6= 1 then it is in contradiction with (O6). So, (1) (t)
W AO cij = W AO
ω (rij , . . . , rij ) (5)
w (x1 , . . . , xn ) = 1 iff xi = 1 for each i =
1, . . . , n, i.e. W AO w satisfies (O3). 2
Step 3 For each alternative aggregate the membership
Corollary 3.1 Let O be a n-dimensional overlap and degrees to each attribute by using W AO w , i.e. for
w = ( n1 , . . . , n1 )T . If O satisfies (O6) then W AO
w is each alternative xi determine the collective overall
a n-dimensional overlap. preference value copi

Proof: Straightforward from Theorem 3.1 W AO w satis- copi = W AO

w (ci1 , . . . , cim ) (6)
fies (O2), (O3), (O4) and (O5). The symmetry, i.e. (O1),
follows from the symmetry of O and the fact that, in this
Step 4 Rank the alternatives in decreasing order with
case, the wi0 s of Eq. (3) are all equals. 2
respect to the collective overall preference values
and choose the alternative with greatest value.
3.2. Multiple attribute group decision making based
on W AOw
3.3. Illustrative example
A solution for a multiple attribute group decision mak-
ing problem (MAGDMP) is a method to choose a We will consider the illustrative example used in [9, 21]
“good” alternative from a set of them, considering a to show how to use our method.
set of attributes for the alternatives and the opinion of Suppose that an investor intends to invest part of his
a group of experts. Formally, in a MAGDMP we have capital in a company. By a market analysis the investor
a finite set X = {x1 , . . . , xn } of feasible alternatives, reduces the spectrum of possible companies into six:
a set A = {a1 , . . . , am } of attributes with their associ-
ated positive weighted vector w = (w1 , . . . , wm )T , and 1. A chemical company, denoted by x1 .
a set d = {d1 , . . . , dt } of decision makers and a positive 2. A food company denoted by x2 .
weighted vector ω = (ω1 , . . . , ωt )T of decision makers 3. A computer company denoted by x3 .
Pk 4. A car company denoted by x4 .
satisfying the usual condition ωi = 1. The method
i=1 5. A furniture company denoted by x5 .
must choose the alternative which “better” satisfies all 6. A pharmaceutical company denoted by x6 .

1475
 Table 1: Assessment of expert e1 .
M (1) a1 a2 a3 a4 a5 a6
x1 0.7 0.8 0.6 0.7 0.5 0.9
x2 0.8 0.6 0.9 0.7 0.6 0.7 Table 4: Standardized decision matrix of expert e1 .
x3 0.5 0.4 0.8 0.3 0.8 0.8 R(1) a1 a2 a3 a4 a5 a6
x4 0.6 0.7 0.6 0.7 0.8 0.6 x1 0.7 0.8 0.6 0.3 0.5 0.1
x5 0.9 0.8 0.4 0.7 0.7 0.8 x2 0.8 0.6 0.9 0.3 0.4 0.3
x6 0.8 0.3 0.7 0.7 0.6 0.7 x3 0.5 0.4 0.8 0.7 0.2 0.2
x4 0.6 0.7 0.6 0.3 0.2 0.4
x5 0.9 0.8 0.4 0.3 0.3 0.2
Table 2: Assessment of expert e2 . x6 0.8 0.3 0.7 0.3 0.4 0.3
M (2) a1 a2 a3 a4 a5 a6
x1 0.6 0.8 0.5 0.6 0.4 0.8
x2 0.7 0.6 0.8 0.6 0.7 0.7
x3 0.7 0.6 0.8 0.7 0.8 0.8
x4 0.6 0.7 0.5 0.6 0.8 0.7
x5 0.7 0.8 0.7 0.7 0.6 0.8
x6 0.6 0.4 0.8 0.7 0.6 0.7

Table 5: Standardized decision matrix of expert e2 .

R(2) a1 a2 a3 a4 a5 a6
The investor is helped by a group of three experts
x1 0.6 0.8 0.5 0.4 0.6 0.2
or decision makers (e1 , e2 and e3 ) with the following x2 0.7 0.6 0.8 0.4 0.3 0.3
weights ω = (0.3, 0.3, 0.4). The group of experts estab- x3 0.7 0.6 0.8 0.3 0.2 0.2
lish that six attributes will be used to evaluate the invest- x4 0.6 0.7 0.5 0.4 0.2 0.3
x5 0.7 0.8 0.7 0.3 0.4 0.2
ments. x6 0.6 0.4 0.8 0.3 0.4 0.3
The benefit attributes are:

a1 ) Benefits in the short term.

a2 ) Benefits in the middle term.

a3 ) Benefits in the long term.

Table 6: Standardized decision matrix of expert e3 .
The cost attributes are: R(3) a1 a2 a3 a4 a5 a6
x1 0.7 0.6 0.6 0.4 0.6 0.3
a4 ) Risk of the investment. x2 0.7 0.6 0.7 0.4 0.4 0.3
x3 0.6 0.5 0.8 0.5 0.2 0.2
a5 ) Difficulty of the investment. x4 0.6 0.7 0.7 0.5 0.2 0.4
x5 0.7 0.8 0.6 0.3 0.4 0.2
a6 ) Other unfavorable factors on the investment. x6 0.4 0.5 0.9 0.3 0.4 0.4

Tables 1, 2 and 3 describes the assesses of the experts

of how much the investment satisfies each attribute, i.e.
they are the decision matrix of each expert.
We normalize these decision matrices resulting in
the standard decision matrices R(1) , R(2) and R(3) de-
scribed, respectively, in Tables 4, 5 and 6. Table 7: Collective decision matrix
In the following, the collective decision matrix from C a1 a2 a3 a4 a5 a6
the standard decision matrices is obtained by using x1 0.703 0.706 0.607 0.412 0.607 0.250
O3 x2 0.746 0.626 0.785 0.412 0.412 0.331
W Aω 0.5M resulting in the matrix of Table 7 (with the x3 0.631 0.534 0.815 0.536 0.229 0.229
values rounded in the third decimal digit). x4 0.626 0.721 0.656 0.460 0.229 0.416
x5 0.768 0.815 0.608 0.332 0.412 0.229
The next step consists in determining the collec-
x6 0.614 0.460 0.845 0.332 0.432 0.383
tive overall preference vector COP by considering
O5
W Aw0.5M where w = (0.1, 0.1, 0.2, 0.2, 0.2, 0.2). The
result is in Table 8.

Table 3: Assessment of expert e3 .

M (3) a1 a2 a3 a4 a5 a6
x1 0.7 0.6 0.6 0.6 0.4 0.7 Table 8: Collective overall preference vector
x2 0.7 0.6 0.7 0.6 0.6 0.7 COP x1 x2 x3 x4 x5 x6
x3 0.6 0.5 0.8 0.5 0.8 0.8 copi 0.658 0.674 0.594 0.621 0.694 0.566
x4 0.6 0.7 0.7 0.5 0.8 0.6
x5 0.7 0.8 0.6 0.7 0.6 0.8
x6 0.4 0.5 0.9 0.7 0.6 0.6

1476
 with a new tool to provide new evidences for the deci-
Table 9: Summary of the ranking obtained in [21], [9]
sion maker can select good alternatives. In fact, in gen-
and with the proposed method.
Methods Ranking
eral, there is no way to validate the quality of a ranking
Maximum x2  x4  x5  x1  x3  x6 and therefore of a method for decision making problems,
Minimum x3 ∼ x5  x1  x4  x2  x6 but when we have several rankings obtained by reason-
NHD x5  x2  x3  x4  x1  x6 able methods, we can determine a good ranking or a par-
WHD x5  x3  x2  x4  x6  x1
Step-IOWAD x5  x4  x6  x2  x3  x1
tial good ranking with the information provided for all
Hurwicz x3  x2  x6  x4  x1  x5 the methods used.
OWAD x5  x3  x2  x4  x1  x6 As a future work we will compare the result with the
AOWAD x5  x2  x4  x3  x6  x1 weighted average generated by the n-dimensional over-
IOWAD x5  x3  x2  x1  x4  x6
AIOWAD x5  x2  x6  x4  x3  x1 lap O0.5 and O2 and the maximum aggregation oper-
Median-IOWAD x5  x1  x6  x2  x4  x3 ator with the generated by other n-dimensional over-
Olympic-IOWAD x5  x4  x1  x2  x3  x6 laps and aggregation functions. In addition, we will
[9] x2 ∼ x3  x6  x4  x1  x5 also study the generation of like OWA operators from n-
O3
W Aω 0.5M x5  x2  x1  x4  x3  x6 dimensional overlaps and apply them in multi-attribute
O3
W Aω 2M x2  x1  x5  x4  x3  x6 group decision making and addapt the method for group
decision making problems based on preference relations
as in [6, 8, 24].
Based on Table 8, we obtain the following rank of the
alternatives:
Acknowledgements

x5  x2  x1  x4  x3  x6 This work was partially supported by the Brazilian

3
O2M
funding agency CNPq – Conselho Nacional de Desen-
If we use the aggregation operators W Aω and volvimento Científico e Tecnológico), under Proc. No.
5 3 5
O2M O2M O2M
W Aw instead of W Aω and W Aw , respectively, 307681/2012-2 and No. 406503/2013-3.
we obtain the following ranking:
References
x2  x1  x5  x4  x3  x6
[1] K. Atanassov and G. Gargov. Interval valued intu-
Observe that this two ranking agree in the last three itionistic fuzzy sets. Fuzzy Sets and Systems, v. 31,
positions, but the alternatives x5 , the winner of the pre- n. 3, p. 343–349, 1989.
vious ranking, now is in the third position. [2] B. Bedregal, G.P. Dimuro, H. Bustince and E. Bar-
On the other hand, the Table 9 adds this two rank- renechea. New Results on overlaps and grouping
ing and the ranking obtained in [9] to the Table 10 of functions. Information Sciences v. 249, p. 148–
[21] which contains the result of twelve methods for 170, 2013.
this same MAGDMP. Analysing this table, we can see [3] G. Beliakov, A. Pradera and T. Calvo. Aggregation
that the fifteen methods return fifteen different rankings. Functions: A Guide for Practitioners, Springer,
Nevertheless, the most (ten) agrees that the best alterna- Berlin, 2007.
tive is x5 and eight that the worst alternative is x6 . So, [4] H.Bustince, J.Fernandez, R.Mesiar, J.Montero and
O3 R.Orduna. Overlap functions. Nonlinear Analysis,
the ranking obtained with W Aω 0.5M seem more reason-
O3 v. 72, p. 1488–1499, 2010.
able than the obtained with W Aω 2M .
[5] H. Bustince, M. Galar, B. Bedregal, SA.
Kolesárová and R. Mesiar. A New Approach to
4. Final remarks Interval-Valued Choquet Integrals and the Problem
of Ordering in Interval-Valued Fuzzy Set Applica-
We propose a method to generate average aggregation tions. IEEE Trans. on Fuzzy Systems, v. 21, n. 6 ,
functions from n-dimensional overlaps and aggregation p. 1150–1162, 2013.
operators, and also propose a method for multi-attribute [6] H. Bustince, M. Pagola, R. Mesiar, E. Hüllermeier
group decision making problems, based on the weighted and F. Herrera. Grouping, Overlap, and General-
average generated by the n-dimensional overlaps O0.5 ized Bientropic Functions for Fuzzy Modeling of
and O2 joint with the aggregation operator “Maximum”. Pairwise Comparisons. IEEE T. Fuzzy Systems, v.
We also consider a multi-attribute group decision mak- 20, n. 3, p. 405–415, 2012.
ing problem which was studied in [21] resulting in 12 [7] S.M. Chen, L.W. Lee, H.C. Liu and S.W. Yang.
differents ranking. The ranking obtained here for O0.5 Multiattribute decision making based on interval-
agrees with the most of them in the best and in the worst valued intuitionistic fuzzy values. Expert Systems
alternative which is just an evidence that this method re- with Applications, v. 39, n. 12, p. 10343–10351,
sult in a reasonable ranking. The goal of introduce this 2012.
new method is not be the best method for MAGDMP, [8] I.A. da Silva, B. Bedregal, R.H.N. Santiago and
neither give the best ranking for a particular example A.D. Dória Neto. A new method for interval-
neither be better than other methods, but just contribute valued intuitionistic group decision making. In:

1477
 Segundo Congresso Brasileiro de Sistemas Fuzzy, guistic information. Omega, v. 40, p. 294–301,
2012, Natal-RN. Recentes Avanços em Sistemas 2012.
Fuzzy. São Carlos-SP: SBMAC, 2012. v. 1. p. 282- [24] D. Paternain, A. Jurio, E. Barrenechea, H.
294. Bustince, B. Bedregal and E. Szmidt. An alter-
[9] I.A. da Silva, B.C. Bedregal, C.G. da Costa, E. native to fuzzy methods in decision-making prob-
Palmeira and M.P. da Rocha. Pseudo-Uninorms lems. Expert Systems with Applications, 39: 7729–
and Atanassov’s Intuitionistic Pseudo-Uninorms. 7735, 2012.
J. of Intelligent and Fuzzy Systems, 2015. doi:
10.3233/IFS-151593.
[10] G.P. Dimuro and B.C. Bedregal. Archimedean
overlap functions: The ordinal sum and the cancel-
lation, idempotency and limiting properties. Fuzzy
Sets and Systems, v. 252, p. 39–54, 2014.
[11] G.P. Dimuro and B.C. Bedregal. On residual impli-
cations derived from overlap functions. Informa-
tion Sciences, 2015. doi:10.1016/j.ins.2015.03.049
[12] G.P. Dimuro, B.C. Bedregal, H. Bustince, M.J.
Asiáin and R. Mesiar. On additive generators of
overlap functions. Fuzzy Sets and Systems, 2015.
doi:10.1016/j.fss.2015.02.008.
[13] G.P. Dimuro, B. Bedregal and R.H.N. Santiago. On
(G,N)-implications derived from grouping func-
tions. Information Sciences, v. 219, p. 1–17, 2014.
[14] B. Feng and F. Lai. Multi-attribute group decision
making with aspirations: A case study. Omega, v.
44, p. 136–147, 2014.
[15] J. Figueira, S. Greco and M. Ehrgott (eds.). Mul-
tiple Criteria Decision Analysis: State of the arts
surveys. Springer, Berlin-Heildelberg, 2007.
[16] D. Gómez, J.T. Rodríguez, J. Montero, H.
Bustince and E. Barrenechea. n-Dimensional over-
lap functions. Fuzzy Sets and Systems, 2015.
doi:10.1016/j.fss.2014.11.023
[17] A. Jurio, H. Bustince, M. Pagola, A. Pradera and
R.R. Yager. Some properties of overlap and group-
ing functions and their application to image thresh-
olding. Fuzzy Sets and Systems, v. 229, p. 69–90,
2013.
[18] E.P. Klement, R. Mesiar and E. Pap. Triangular
Norms. Kluwer Academic Publishers, Dordrecht,
2000.
[19] N. Madrid, A. Burusco, H. Bustince, J. Fernan-
dez and I. Perfilieva. Upper bounding overlaps by
groupings. Fuzzy Sets and Systems, v. 264, p. 76–
99, 2015.
[20] J.M. Merigó. A unified model between the
weighted average and the induced OWA operator.
Expert Systems with Applications, v. 38, n. 9, p.
11560–11572, 2011.
[21] J.M. Merigó and M. Casanova. Decision-Making
with distance measures and induced aggregation
operators. Computer & Industrial Engineering,
v. 60, p. 66–76, 2011.
[22] S.M. Mousavi, F. Jolai and R. Tavakkoli-
Moghaddam. A Fuzzy Stochastic Multi-Attribute
Group Decision-Making Approach for Selection
Problems. Group Decision and Negotiation, v. 22,
n. 2, p. 207–233, 2013.
[23] J. Pang and J. Liang. Evaluation of the results
of multi-attribute group decision-making with lin-

1478